Add property, fuzz, concurrency, signal, and performance test suites

[codeforester]

Parent: #50

Train position: 18 of 22
Phase: P3 — Ecosystem trust

Problem

The highest-risk parsing, persistence, concurrency, interruption, and startup paths need adversarial and performance regression coverage.

Acceptance criteria

• Property and fuzz tests exercise argv/history redaction and command-protocol framing/validation.
• Concurrency and multiprocessing tests stress history, logging, atomic metadata, plugin discovery caches, and retention.
• Signal tests cover Ctrl+C/Abort and applicable termination behavior without leaked contexts, files, or handlers.
• Benchmarks track import/startup time and per-invocation filesystem overhead with documented budgets.
• Seeded reproducers and platform-specific expectations keep failures deterministic.

Dependencies

• Finalize run metadata and normalize every invocation outcome #52
• Harden sensitive-option redaction and runtime cleanup boundaries #54
• Implement atomic run bundles and complete-bundle retention policies #62
• Define stable structured-log, JSON output, and JSON error contracts #63